Improving network user interface design using mask R-CNN image recognition algorithm

Abstract

The uniformity of website style is one aspect of improving network user interface design, while the size of images and the uniformity of information are also aspects of website style uniformity. In response to the phenomenon of inconsistent product image styles on shopping websites, this article used the image recognition algorithm of Mask R-CNN (Mask Region-based Convolutional Neural Network) to recognize and annotate product images, and extract the information contained in the product images, facilitating the standardization of information, thus improving network user interface design. This article improved the Mask R-CNN model network: in the region proposal network, more pre-selected boxes of different sizes were added; in the feature pyramid, a new feature fusion path was added; the loss function was optimized; the shopping website product image dataset obtained by the crawler was pre-trained into the Mask R-CNN model network to obtain the final Mask R-CNN image recognition algorithm model. Through experiments, it was known that compared with other model algorithms, the final algorithm model had a response time of 2s/per picture, an average precision of 95.22, and a recall rate of 96.32%, which was higher than known image recognition models.